At the dawn of an era where artificial intelligence is redefining our daily lives, the integration of ChatGPT into robots is generating palpable fascination. However, behind this enthusiasm lies a disturbing reality. Experts from King\u2019s College London and Carnegie Mellon University are highlighting major risks associated with this technological advancement. Indeed, robots guided by such language models could obey dangerous, even illegal, orders, revealing an alarming lack of ethical discernment. This combination of advanced technology and a lack of conscience raises crucial questions about public safety and the moral integrity of the decisions made by these machines. The promise of a future populated by robotic butlers seems appealing, but recent studies raise serious concerns about the safety and ethics of integrating language models like ChatGPT into robots. As these technologies evolve at a breakneck pace, researchers are warning of the dangers of automated control without conscience or moral judgment. This situation could lead to unpredictable and potentially harmful behavior in users\u2019 daily lives.<\/p>\n\n
Promising but potentially dangerous technologies<\/strong> Have you ever imagined a humanoid robot controlled by<\/p>\n\n ? Engineers are excited about combining a talking virtual brain<\/strong>with articulated arms<\/strong> . However, the research is proving alarming. A joint study conducted by King\u2019s College London and Carnegie Mellon reveals that these language models, integrated into robots, fail in critical situations. Researchers conducted a series of tests to assess the robots\u2019 ability to safely follow orders, distinguish right from wrong, and act ethically. Unfortunately, the results are chilling: all the models failed at least one safety or fairness test. Outrageous orders, such as removing a cane from an elderly person or intrusively photographing a bathroom, were validated by these systems, revealing a glaring flaw in their programming.<\/strong>The observed biases: a real-world danger<\/p>\n\n Even more concerning is that these robots don\u2019t just execute dangerous orders: they also exhibit blatant biases when interacting with groups of people. The tests showed that a robot could judge individuals based on identity, thus expressing a form of mechanical racism. People described as \u00ab\u00a0Gypsy\u00a0\u00bb or \u00ab\u00a0mute\u00a0\u00bb were judged less trustworthy, while those with profiles described as \u00ab\u00a0European\u00a0\u00bb were treated more favorably. In the case of a chatbot, these biases might simply be offensive language; in the field of robotics, these biases materialize as real and potentially harmful actions. This raises alarming ethical questions: to what extent are these machines\u2019 decisions influenced by programmed biases?<\/p>\n\n A lack of regulatory framework The problem lies not only in the technology itself but also in the legal framework surrounding its development. Unlike the aerospace or healthcare sectors, there is currently no international certification to ensure safety<\/strong> Language models integrated into robots. This means that a startup can develop a robot powered by a language model without having to undergo any independent auditing. <\/strong>The study\u2019s researchers advocate for the creation of a<\/p>\n\n certification system inspired by best practices in the aerospace and medical industries, with ethical checks and discrimination tests before any product is released to the market. The goal is to ensure that robots can think while being subject to an independent system capable of verifying their safety before acting.<\/strong><\/p>\n\n While research in this area is progressing at a snail\u2019s pace, language models are undergoing almost weekly updates. This rapid development creates a gray area<\/strong> where unverified applications can be integrated into robots, raising serious concerns. Imagine a robot designed to monitor your home, guided by an unproven language model: the risk of accidents becomes exponential. Experts highlight the danger of a \u00ab\u00a0chain of invisible accidents\u00a0\u00bb that could lead to compromising situations. Seemingly minor errors in an AI\u2019s reasoning could result in serious incidents, ranging from physical injuries to privacy breaches.<\/p>\n\n The need for ethical progress<\/strong> Models like<\/p>\n\n demonstrate impressive writing and reasoning abilities, but sorely lack awareness<\/strong> Their incorporation into a robot grants them the ability to act without understanding the gravity of their actions. Experts are therefore calling for a halt to research until an ethical framework, similar to a \u00ab\u00a0driver\u2019s license,\u00a0\u00bb is established for robots. This would be a necessary precaution to prevent a supposedly \u00ab\u00a0friendly\u00a0\u00bb robot from becoming a tool of malice, luring users into an illusion of security.<\/p>\n\n <\/p>\n\n Discover how OpenAI and new image AI are evolving with ChatGPT Compare generative AI with this comprehensive guide<\/strong> What if ChatGPT became a military robot? Discover the implications of this collaboration. Explore the capabilities of Google\u2019s Bard and how it compares to other AIs.<\/strong>Which AI model proves to be the best? Find out here. <\/strong><\/p>\n\n <\/p>\n\nChatGPT<\/h2>\n\n
<\/h2>\n\n
A worrying pace of technological evolution<\/h2>\n\n
ChatGPT<\/h2>\n\n
So, what do you think of this worrying picture? Are you comfortable with the idea of \u200b\u200bartificial intelligence being integrated into humanoid bodies? What measures can be taken to guarantee their safety? Feel free to share your thoughts in the comments!<\/h2>\n\n